Nice commentary, I also read Mallory’s paper and wondered what you would say about the idea of using deep neural nets as measurement tools in science, more specifically.
I think things are generally headed in this direction, i.e., that next-generation measurement tools in scientific applications will be incorporating neural nets and other computational transformations. That said, one should distinguish neural nets as function approximators from neural nets as inferential devices. I can envision both roles as modules in measurement tools, but they should be treated differently.
Nice commentary, I also read Mallory’s paper and wondered what you would say about the idea of using deep neural nets as measurement tools in science, more specifically.
I think things are generally headed in this direction, i.e., that next-generation measurement tools in scientific applications will be incorporating neural nets and other computational transformations. That said, one should distinguish neural nets as function approximators from neural nets as inferential devices. I can envision both roles as modules in measurement tools, but they should be treated differently.